performance prediction of conventional and modified solar stills using levenberg marquardt algorithm-based artificial neural network model: an experimental and stochastic evaluation

A neural network (ann) model employing the levenberg-marquardt (lm) algorithm was formulated and employed to study the functionality of both conventional (css) and modified (mssw and mssu) solar distillation systems. numerous input factors, comprising solar irradiance, wind speed, atmospheric conditions, glass properties, and water temperatures, were carefully selected, with the yield of distilled water serving as the target variable. the model underwent a process of testing, training, and validation utilizing empirical data obtained from css, mssw, and mssu setups, achieving a confidence level of 95%. after validation, the model's capabilities were utilized to forecast the distilled water output based on a distinct set of input parameters. the outcomes unveiled a negligible deviation, with a maximum disparity of 3.1% and 4.6% observed in comparison to the experimental findings for mssw, and mssu setups, respectively, thereby signifying a substantial agreement between theoretical predictions and experimental observations. furthermore, the model exhibited outstanding accuracy in contrast to well-established numerical models proposed by several researchers, thereby demonstrating its efficacy in predicting the performance of solar stills.